What are sensitivity and specificity?
Sensitivity and specificity are two measures of how accurate a medical test is. Sensitivity is the share of people with a condition whom the test correctly flags. Specificity is the share of people without it whom the test correctly clears.
Also known as: true positive rate, true negative rate
Researched and fact-checked by AI, with no human review. 6 sources listed below. How we verify
Last updated
How it works
Both measures compare a test's answers with a reference standard, meaning the best method available for determining whether the condition is present or absent. The U.S. Food and Drug Administration (FDA) issued statistical guidance in March 2007. It defines sensitivity as the proportion of people with the target condition who test positive. It defines specificity as the proportion of people without it who test negative.
The FDA recommends reporting the two together, with 95 percent confidence intervals. It notes that estimates for the same test can differ from study to study depending on the types of people included.
Why it matters
The two measures can pull against each other. A 2017 article in Frontiers in Public Health cites a study in which a test showed 100 percent specificity but only 45.2 percent sensitivity. It says this indicated the cutoff might have been too strict. The article also argues that the pair describe the test, not the patient. What a positive result means for one person is a separate figure, the positive predictive value: the share of people testing positive who have the condition.
A draft FDA guidance on AI-enabled devices was issued in January 2025 and still listed as a draft in September 2026. A fictitious worked example in it shows the gap. A model with 84 percent sensitivity and 83 percent specificity is used on 1,000 patients, 20 percent of whom have the disease. It correctly flags about 168 of those 200 patients and wrongly flags about 136 of the other 800. So about 55 percent of its positive results are right.
Where things stand in 2026
IDx-DR is an AI system that checks retinal images for diabetic retinopathy. A company-funded pivotal trial of it enrolled 900 people at 10 U.S. primary care sites in 2017. The system showed 87.2 percent sensitivity and 90.7 percent specificity, above preset thresholds of 85 and 82.5 percent. The FDA authorized it in 2018.
A Swedish randomized trial of 105,934 women was published in The Lancet in January 2026. It found that AI-supported mammography screening had a sensitivity of 80.5 percent, compared with 73.8 percent for standard double reading without AI. Specificity was 98.5 percent in both groups.
The FDA draft guidance says labeling for an AI-enabled device should give performance estimates with confidence intervals and explain performance across subgroups such as sex, age and race.
Sources
- Statistical Guidance on Reporting Results from Studies Evaluating Diagnostic Tests - Guidance for Industry and FDA Staff, U.S. Food and Drug Administration
- Sensitivity, Specificity, and Predictive Values: Foundations, Pliabilities, and Pitfalls in Research and Practice, Frontiers in Public Health, via PubMed Central (US National Library of Medicine)
- Artificial Intelligence-Enabled Device Software Functions: Lifecycle Management and Marketing Submission Recommendations (Draft Guidance, issued January 7, 2025), U.S. Food and Drug Administration
- Artificial Intelligence-Enabled Medical Devices, U.S. Food and Drug Administration
- Pivotal trial of an autonomous AI-based diagnostic system for detection of diabetic retinopathy in primary care offices, npj Digital Medicine, via PubMed Central (US National Library of Medicine)
- Interval cancer, sensitivity, and specificity comparing AI-supported mammography screening with standard double reading without AI in the MASAI study: a randomised, controlled, non-inferiority, single-blinded, population-based, screening-accuracy trial, The Lancet, abstract via PubMed (US National Library of Medicine)
